A neural network thrust force controller to minimize delamination during drilling of graphite-epoxy laminates

Delamination is a well-recognized problem associated with drilling fiber-reinforced composite materials (FRCMs). The most noted problems occur as the drill enters and exits the FRCM. Since drilling is often a final operation during assembly, any defects introduced in parts through the drilling process that result in the part being rejected represent an expensive loss. Studies based on linear-elastic fracture mechanics theory have proposed critical cutting and thrust forces in the various drilling regions that can be used as a guide in preventing crack growth or delamination. Using these critical force curves as a guide, a thrust force controller was developed to minimize the delamination while drilling a graphite-epoxy laminate. A neural network control scheme was implemented which required a neural network identifier to model the drilling dynamics and a neural network controller to learn the relationship between feed rate and the desired thrust force. Experimental results verifying the validity of this control approach as well as the robustness of the design are presented. Visual measurements of the delamination zones were used to quantify the benefits of the thrust force controlled drilling process versus the conventional constant feed rate drilling process. Copyright (C) 1996 Published by Elsevier Science Ltd